LiDAR SLAM for Mobile Robot

Abstract

Simultaneous localization and mapping (SLAM) means that a mobile device starts from a location in an unknown environment, observes its own position, pose, and motion trajectory through sensors (e.g., LiDAR, camera), and constructs an incremental map based on its own position. LiDAR SLAM technology has become the core technology for autonomous localization and navigation of robots due to its performance advantages such as stability and reliability. This chapter covers the principle and usage of 2D LiDAR SLAM algorithms, which scan the environment profile using LiDAR for localization and mapping. Common algorithms include Gmapping, Hector SLAM, Karto SLAM, and Cartographer. Based on the understanding of 2D LiDAR algorithms, readers can learn to use 3D multi-line LiDAR SLAM algorithms, such as LOAM, LeGO-LOAM, LIO-SAM, and Fast-LIO.

Appendices

Further Reading

Ji Zhang and Sanjiv Singh proposed LiDAR odometry and mapping (LOAM) in real-time using range measurements from a 2-axis LiDAR moving with 6-DOF¹. This is a difficult issue because the range measurements are received at different times, and errors in the motion estimation can cause the misregistration of the resulting point cloud. Presently, coherent 3D maps can be built by offline batch methods, which often employ loop closure to correct for drift over time. This method achieves both low-drift and low-computational complexity without the requirement for high accuracy ranging or inertial measurements. The key concept for obtaining this level of performance is dividing the complex problem of SLAM, which seeks to optimize a large number of variables simultaneously, using two algorithms. One algorithm performs odometry at a high frequency but low fidelity to estimate velocity of the LiDAR. Another algorithm runs at a frequency of an order of magnitude lower for fine matching and registration of the point cloud. Combining the two algorithms enables the method to realize real-time mapping. The method has been evaluated by a large set of experiments as well as on the KITTI odometry benchmark. The results indicate that the proposed method can achieve accuracy at the level of state-of-the-art offline batch methods.

Tixiao Shan and Brendan Englot proposed a lightweight and ground-optimized LiDAR odometry and mapping method (LeGO-LOAM)², to obtain real-time six degree-of-freedom pose estimation for ground vehicles. LeGO-LOAM is lightweight, and it can achieve real-time pose estimation on a low-power embedded system. LeGO-LOAM is ground-optimized, as it leverages the presence of the ground plane in its segmentation and optimization steps. In LeGO-LOAM, point cloud segmentation is first applied to filter out noise, and feature extraction is utilized to obtain distinctive planar and edge features. A two-step Levenberg–Marquardt optimization method is subsequently applied to the planar and edge features to solve different components of the six degree-of-freedom transformation across consecutive scans. Using datasets gathered from variable terrain environments with ground vehicles, to compare the performance of LeGO-LOAM with LOAM, experimental results showed that LeGO-LOAM achieves similar or better accuracy with reduced computational expense. LeGO-LOAM was integrated into a SLAM framework to eliminate the pose estimation error caused by drift, which was tested using the KITTI dataset.

Exercises

• 1. Type the start command in the terminal: roslaunch turtle_tf2 turtle_tf2_ demo.launch

  This command is to start the small turtle following program built in ROS, the movement of one turtle can be controlled by keyboard, and meanwhile, the other turtle can follow the movement.

• 2. Some practical tools built in ROS can easily and quickly implement a function or debug a program. During the operation of the turtle following in exercise (1), please use the rosbag to record the movement, and play back the recording at 0.5 times speed.

• 3. What are the optimization algorithms based on the back-end processing of the SLAM system? What are the advantages and disadvantages of different algorithms?

• 4. Using an example to describe the TF from base_link to base_laser of a mobile robot.

• 5. Gmapping is one of the classic 2D laser SLAM algorithm, what is the core idea and process of the Gmapping algorithm?

• 6. What are the inputs and outputs of the ROS Gmapping package? What topics are subscribed and published? Please install Gmapping package and dependencies, download the relevant data set and replay it, and then, run the Gmapping algorithm to build a map.

• 7. Describe the principles, advantages, and disadvantages of the Hector SLAM algorithm.

• 8. Write the mathematical expression for the Gauss–Newton method.